Susumu, You have got the correct top level idea. Though glusterhpc can do a few more things. It can not only replicate master node to all slaves, but it can also pick one of the client nodes as the image source too. It not only replicates the system image, but also installes a bunch of HPC apps extra (like mpich, lam, slurm etc) which you can chose during installation and pre-configures them for straight away use. For glusterEP, your understanding is correct. What is more is, since we have authentication/encrypted transfer, it can also be used for corporate backup where users can capture their system state safely in the master node and others cannot access it without password. We also add more advanced features like per-user quota and suppport for Microsoft Windows in the coming releases of GlusterEP regards, avati On Wed, Jun 28, 2006 at 03:13:48PM -0700, Susumu Urata (Technocross) wrote: > I would like to confirm the following. > > 1. GlusterHPC objective is to distribute the master image to all > the clients so that the each computing node has the same image to > perform the HPC jobs. > This is why the Gluster collects the client network information and > then distributes the image to everyone at the same time. Once the > image is distributed, the master and the clients can be booted up > normally. > > 2. GlusterEP objective is to capture the individual client node > image and store it to the master node so that the each client node > can be rebooted with Gluster client to provision the system whenever > the local boot image is lost or add a new node. Provisioning allows > the stored image to be loaded to the local disk based on the > username. This way, the client node images can be managed centrally > and no more loading of OS with a CD is required. > > Susumu Urata > Technical Consultant > > 408-873-2760 (Voice) > 408-873-2761 (Fax) > > > > _______________________________________________ > Gluster-devel mailing list > Gluster-devel@xxxxxxxxxx > http://lists.nongnu.org/mailman/listinfo/gluster-devel >
